A look at the technology of the future – Chatbot

A look at the technology of the future – Chatbot

Thanks to ChatGPT, natural language processing has reached an unprecedented level of quality, opening up entirely new possibilities in human-machine communication.

GPT-3, ChatGPT, and OpenAI are currently riding a wave of popularity on the web. The latest version of the chatbot called ChatGPT, created by the American company OpenAI from San Francisco, is causing quite a media buzz, delighting users from various fields with its capabilities. This technological milestone in text recognition using artificial intelligence (AI) could also create a new perspective for logistics.

As the name suggests, a chatbot is a combination of the words “chat” and “robot”. It can be a text or voice dialogue system that enables real-time communication between humans and machines based on natural language. Simple chatbots search the web or a defined dataset to find keywords entered by the user. They then use a programmed decision tree to select predefined texts or text modules from a database of possible answers, allowing them to provide them as solutions. Websites often contain these types of basic chatbots to provide quick and easy help to the user.

Smarter chatbots use natural language processing (NLP) to determine the context for an entirely new written or spoken sentence. The goal is to find appropriate answers based on a knowledge base. NLP is based on artificial intelligence (AI) and machine learning algorithms. Until now, the quality of results obtained from such intelligent chatbots had limited practical application. As tests at DACHSER have shown, available algorithms from well-known cloud solution providers correctly recognized, at best, the context of 60 to 80 percent of input data.

GPT-3 AI Algorithm

Now, OpenAI has reached a new dimension of context recognition with its ChatGPT software. GPT stands for “generative pre-trained transformer” and refers to a special new type of NLP model that the company developed in 2018. It has now released GPT-3, the third version of the NLP model, which is the foundation of ChatGPT’s success. In the meantime, the company has moved away from its original open-source approach and is adopting a more commercial tactic.

Unlike previous chatbots, ChatGPT is capable of providing natural-sounding conversations, enriched with a wealth of factual information, even over multiple rounds of text input. This has been confirmed by numerous studies, as well as by preliminary tests at DACHSER. Based on a deep learning model and a large training dataset, ChatGPT is able to handle all kinds of questions across a wide range of topics. Its answers are similar in pattern and structure to human speech. However, it cannot (yet) replace independent human thinking and critical probing of content and context. When it comes to mathematical skills and certain logical tasks, ChatGPT also still has a long way to go. But in semantics, this is not the case: every chatbot response is always built word by word by calculating the probability for the next word. Furthermore, ChatGPT can generate or correct software code. Writing messages, poems, summaries, or market analyses in various languages is also basically no challenge for the ChatGPT algorithm.

Artificial intelligence also has applications in other areas. Beyond text and software code, it can, for example, generate images, videos, or music. Chatbots with a level of context recognition similar to ChatGPT will also have an impact on logistics and IT. Automating communication with machines, systems, and vehicles, as well as with customers, partners, and employees, could make this technology have an unprecedented share in daily operations.

In addition to a very large training dataset, a special feature of GPT-3 is the way the underlying AI model was trained. In the case of ChatGPT, forms of supervised learning and reinforcement learning were combined, and humans themselves were involved in the training process through active learning. Simply put – during the supervised learning process, humans served as trainers, taking on the roles of both the questioner and the responder. In the reinforcement learning phases, trainers developed a method for ranking the quality of the chatbot’s responses to a conversation that had already taken place. Based on these rankings, they created reward models, which were then incorporated into subsequent training iterations.

ChatGPT also contains a rule-based mechanism designed to prevent inappropriate responses on a given topic. The lack of this capability was the main criticism of the predecessor to GPT-3 and comparable models. However, even ChatGPT still struggles to reliably exclude plagiarism, fake news, and discriminatory or sexist statements. The chatbot is growing as a writing assistant – but it is also controversial.

Communication Automation

One thing is certain: ChatGPT, along with its underlying technology, GPT-3, is one of the hottest tech topics of the year. It will not solve all tasks and problems, but ChatGPT has the potential to generate new innovative processes, products, and business models. You can try it out yourself at https://chat.openai.com

Source: Dachser.